In other developments, Lyft has unveiled a new collaboration set for 2026 involving Holon autonomous shuttles powered by Mobileye Drive. While such announcements are common and typically not headline-worthy, they offer a useful lens through which to compare the progress of various players in the self-driving space.

During Tesla’s recent earnings call, the company revealed that its robotaxis in Austin have collectively traveled approximately 7,000 miles. With around a dozen vehicles reportedly active over the past 30 days, this averages to fewer than 20 miles per vehicle per day—an unexpectedly low figure. Although Ashok Elluswamy claimed there were “no notable safety-critical interventions,” this assertion conflicts with accounts from some riders. One influencer reported that a Tesla robotaxi struck a Toyota’s wheel in a parking lot, while multiple videos show the vehicles dropping passengers in hazardous spots. Additionally, there have been instances of unsafe or illegal maneuvers that didn’t result in collisions—technically not “interventions” under Tesla’s definition, but still concerning.

There are many types of interventions, and some companies argue that if no damage occurs, no real issue exists. In Q1, Elon Musk claimed Tesla was achieving roughly 10,000 miles per intervention, though he didn’t specify the severity. Given the number of incidents observed in just 7,000 miles, that claim seems optimistic. We’re not seeing footage of every mile driven, and influencers given access often don’t document or highlight problems. To meet Musk’s stated goal of being “much better than a human,” Tesla would need to go millions of miles without any serious safety incident—a bar far from being met.

The minimal daily usage strongly suggests Tesla is exercising extreme caution in deploying its service, which is the appropriate safety posture for a system still in early development.

Tesla has expanded its operational zone to resemble a massive upside-down Tesla logo (or perhaps something else when viewed right-side up), and they claim another expansion is imminent. This makes the low mileage even more puzzling. However, expanding service areas isn’t particularly difficult for a supervised system like Tesla’s. In fact, Tesla could theoretically roll out coverage across much of the country in this mode—if they were willing to staff the necessary remote monitoring teams. A supervised system can operate even if it’s 100 times less capable than what’s needed for full autonomy, which appears to be where Tesla currently stands relative to Musk’s own benchmarks. Therefore, his boast that Tesla will cover a larger area than competitors isn’t a meaningful comparison. The company also expressed optimism about serving "half the population of the USA by the end of this year"—an ambitious-sounding goal, but not especially bold for a supervised service. Some states may require the safety operator to sit in the driver’s seat rather than the passenger seat, depending on local regulations. Interestingly, today there are already Tesla owners using FSD and Autopilot while giving rides via Uber—something that appears legally permissible, though not officially sanctioned.

On a positive note, early reports suggest Tesla has made notable improvements to its pick-up functionality, a critical aspect of any robotaxi service. Initial videos showed riders walking long distances in parking lots, indicating the system wasn’t yet capable of handling curbside pickups involving waiting. (Drop-offs are simpler since they require little to no waiting.) If these improvements hold, it would reflect a significant leap in the system’s adaptability and real-world performance.

Misleading Chart

Tesla also highlighted its latest Autopilot safety report during the call, sharing a chart on X/Twitter:

This chart is misleading. As noted in the fine print, the blue line defines “accidents” as airbag deployments, and applies to Autopilot usage—which occurs 90% on freeways, where crash rates per mile are about one-third of those on all roads. Furthermore, the data for Autopilot human driving comes from Tesla owners, who tend to be wealthier, older, and statistically safer drivers than the general population.

According to State Farm, the average driver experiences an airbag deployment about once every 4 million miles, though this varies by manufacturer due to differing airbag deployment thresholds. When all these factors are considered, the actual safety benefit of Autopilot driver is likely negligible—and possibly slightly worse than human-only driving. Yet Tesla presents it as 10 times safer than the orange line, which represents police-reported crashes (regardless of airbag use) across all drivers, vehicles, and road types. This portrayal is clearly inaccurate. While the safety numbers have shown modest improvement, the core issue remains: Tesla possesses the real data but chooses not to disclose it. Transparency would go a long way. In the past, Tesla at least disclosed the data discrepancies in footnotes without making overt false claims; this time, the implication is more aggressively misleading.

Tesla reiterated its confidence in achieving major milestones by 2025 and 2026, while acknowledging they sometimes miss their timelines. Meanwhile, Mobileye has been making similar claims, asserting that its self-driving technology will launch through partners in 2026. This includes earlier plans for a robotaxi service with MOIA and Volkswagen, and today’s announcement from Lyft that it will offer autonomous rides using shuttles from Benteler’s Holon division. Holon is already operating self-driving shuttles in Jacksonville, Florida.

It remains unclear how Holon’s fixed-route, shared-ride model will integrate into the Lyft app, which currently doesn’t support booking such services. Lyft may adapt its platform to allow point-to-point trips or modify the rider experience. The company is keen to match Uber’s rapid pace of autonomous partnerships.

What’s less certain is whether Mobileye Drive—the company’s fully unsupervised self-driving system—will be ready in time. Mobileye has a strong team and leads the ADAS market for traditional automakers, but we’re now entering the “Yoda phase” of autonomous driving: “there is no try, only do.” Early reports from MOIA’s pilot in Germany suggest safety drivers are still frequently intervening. If Mobileye and its partners hope to convince the public they can operate safely without human oversight by 2026, they’ll need to demonstrate concrete, verifiable progress—soon.

The above is the detailed content of Tesla Robotaxis Only Go 20 Miles/Day. Meanwhile Where's Mobileye?. For more information, please follow other related articles on the PHP Chinese website!